master-thesis
📄 My Master's Thesis in Computer Science, written at the Norwegian University of Science and Technology (NTNU)
Science Score: 26.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (8.5%) to scientific vocabulary
Keywords
Repository
📄 My Master's Thesis in Computer Science, written at the Norwegian University of Science and Technology (NTNU)
Basic Info
- Host: GitHub
- Owner: andstor
- Language: TeX
- Default Branch: master
- Homepage: https://hdl.handle.net/11250/3015521
- Size: 70 MB
Statistics
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
- Releases: 0
Topics
Metadata Files
README.md
Secure Smart Contract Code Synthesis with Transformer Models
Abstract
Writing Smart Contracts (SCs) are hard. Writing secure SCs is even harder. Automatic code generation is by many considered the holy grail in the field of computer science. Recent advances in transformer models have shown great potential in the area of synthesizing code from code comments. However, it is just as important how these models are best put to use. In this thesis, it is investigated: how to automatically generate Smart Contract code with transformer-based language models, by inputting comments to guide the code generation? Due to the monetary and immutable nature of blockchain, security in SCs is of uttermost importance. This thesis also investigates: how to generate secure Smart Contract code with transformer-based language models? A design science research approach is adopted for answering these research questions. For automatically synthesizing SC code, the 6 billion parameter model GPT-J by EleutherAI is fine-tuned on SC code. For this task, the currently largest dataset of real SC code is constructed, containing 186,397 contracts. To evaluate the comment-aid approach to generate code, the original code was used as the ground truth. This code was then compared with the generated code, and their differences were measured using the BiLingual Evaluation Understudy (BLEU) score. Results of the evaluation show that this approach results in a BLEU score of 0.557, which is beyond the state-of-the-art. For generating secure Smart Contract code with transformer-based language models, a novel method named security conditioning is proposed. From both automatic and manual evaluation of the technique, the evaluation results show that security conditioning produced secure code.
Building document locally
A Makefile is provided for building the document locally. This requires a LaTeX compiler, such as texlive, installed locally, which has to provide the commands pdflatex and biber.
FAQ
If sources produces an error, try running
rm -rf `biber --cache`
Owner
- Name: André Storhaug
- Login: andstor
- Kind: user
- Location: Trondheim 🇳🇴
- Company: NTNU
- Website: https://andre.storhaug.no
- Repositories: 87
- Profile: https://github.com/andstor
🎓 CS PhD student @ Norwegian University of Science and Technology (NTNU)
GitHub Events
Total
Last Year
Issues and Pull Requests
Last synced: 6 months ago
All Time
- Total issues: 0
- Total pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Total issue authors: 0
- Total pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0
Past Year
- Issues: 0
- Pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Issue authors: 0
- Pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0